Research on the Schedulability Prediction Method of Electric Vehicle Cluster Based on CNN-BiLSTM-Attention and Model Coupling Drive

Under the background of dual-carbon, aiming at the difficulty of flexible scheduling and the accuracy of schedulability prediction of massive electrical vehicle (EV) after grid connection, a CNN-BiLSTM-Attention and model coupling driven schedulability prediction method considering the stopping characteristics of EV cluster is proposed. Firstly, the convolutional neural network (CNN) is used to fully extract the detailed features of EV stopping data, and then the bidirectional long short-term memory (Bi-LSTM) network is cascaded to bidirectionally mine the spatio-temporal characteristics of EV stopping data, and the attention mechanism is integrated to dynamically allocate weights for the stopping features to improve the prediction accuracy of EV stopping parameters. Secondly, considering the natural aggregation of charging stations, an EV cluster is constructed by aggregating individual EVs based on Minkowski summation, and it is simplified as a generalized energy storage (GES) device. Finally, the capacity and power boundary of the GES model are described according to the prediction results of the EV cluster stop parameters, which can effectively take into account the user's energy consumption law and quantify the EV cluster schedulability. The simulation results show that compared with the single-driven schedulability prediction method, the combined method can further clarify the energy consumption behavior of EV users, reduce the model dimension, and realize the accurate description of the schedulability of EV clusters.